Time-Efficient Adaptive Segmentation Algorithm forIC Layers Images

Abstract

In this paper we introduce adaptive threshold algorithm for grayscale images which substantially enhances object extraction precision while maintaining robust calculation performance even for very large images. The algorithm was developed for feature analysis in low quality integrated circuit images obtained using optical microscopy equipment. Our proposed adaptive segmentation algorithm gives standard deviation 5 times smaller than ordinary bi-level segmentation algorithm when comparing segmented object contour differences to manually segmented image. This technique has been applied in industry.